NEP 29 - drop Python 3.6 support?

Hey all,

I noticed at Dask’s community issues repo that starting next week, Numpy’s NEP 29 recommends starting to drop support for Python 3.6. We’ve already decided to go with NEP 29 once before.

This would mean, skimming the seemingly relevant to us issues from the changelog, that:

P.S. Dask’s community repo is a neat idea - it’s basically this Discourse forum, but as Github issue tracker.

I’d be okay with dropping Python 3.6 support anytime. I think we should feel free to use features from Python 3.7, but keep supporting Python 3.6 in the meantime. Pretty much…we should drop 3.6 support right up until we implement a Python 3.7 feature for the first time. But if you’re excited to use a dataclass, go for it!

1 Like

One other thought…PyPy currently implements Python 3.6.9, though they are likely to support Python 3.7 soon. If we’d like to try supporting PyPy, then we should stick with Python 3.6 for a while longer.

I don’t think pypy really caught on in the scientific Python. I haven’t, at least, heard of it being used a lot.